hadoop-mapreduce-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Konstantin Boudnik (JIRA)" <j...@apache.org>
Subject [jira] Commented: (MAPREDUCE-1933) Create automated testcase for tasktracker dealing with corrupted disk.
Date Fri, 30 Jul 2010 01:53:17 GMT

    [ https://issues.apache.org/jira/browse/MAPREDUCE-1933?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12893905#action_12893905

Konstantin Boudnik commented on MAPREDUCE-1933:

- I still see that string "mapred.local.dir" is used all over the place in the trunk patch.
First, don't use string literals where there's a named constant for the same; Second: this
is a deprecated parameter: javac will curse and raise the level of warnings.
- Shall these two be swapped, actually?
+    cluster.tearDown();
+    cluster.restart();
Perhaps, cleaning the folders should also be done before the {{tearDown()}} call?
-  Is there defined constants for something like "test.randomwrite.bytes_per_map" ? If not
- define a private member of the test class instead of using string literals.

> Create automated testcase for tasktracker dealing with corrupted disk.
> ----------------------------------------------------------------------
>                 Key: MAPREDUCE-1933
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-1933
>             Project: Hadoop Map/Reduce
>          Issue Type: Test
>          Components: test
>            Reporter: Iyappan Srinivasan
>            Assignee: Iyappan Srinivasan
>         Attachments: 1933-ydist-security-patch.txt, 1933-ydist-security-patch.txt, MAPREDUCE-1933.patch,
MAPREDUCE-1933.patch, MAPREDUCE-1933.patch, TestCorruptedDiskJob.java
> After the TaskTracker has already run some tasks successfully, "corrupt" a disk by making
the corresponding mapred.local.dir unreadable/unwritable. 
> Make sure that jobs continue to succeed even though some tasks scheduled there fail.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message